Experimental and Numerical Investigation of Fluid Flow

and Mixing in Pachuca Tanks

E. RODRIGUEZ M., A.H. CASTILLEJOS E., and F.A. ACOSTA G.
This study reports laboratory and computational work carried out to determine the eect of
operating and design parameters on the motion of the liquid and the mixing of a solute in
Pachuca tanks. In the laboratory tanks, the liquid velocity eld was measured using particle
image velocimetry (PIV) and mixing was characterized using a stimulus-response technique.
Visualization of the air-water ow suggested the suitability of a one-phase variable density
turbulent recirculating ow model coupled to the drift ux model, to describe the motion of the
water phase and the gas holdup, in two dimensions (2-D) and in steady state. The dispersion of
the solute tracer was simulated by solving the unsteady state turbulent mass transfer equation in
three dimensions (3-D). The computational predictions give a good estimate of laboratory
measurements of the inuence that operating and design variables have on liquid circulation
velocity, liquid ow pattern, gas holdup, and mixing evolution and time. Fluid ow simulation
of industrial-scale tanks revealed that the recirculating loop that forms in their annular section is
more intense and extends over a larger proportion of the reactor height as the draft tube/tank
diameters ratio, dd/dt, decreases, at the same time the supercial liquid velocity in the draft tube
increases. These features suggest that tanks with dd/dt~ 0.1 promote conditions for good particle
suspension by hindering the settling of particles in the annulus and favoring their lifting in the
draft tube; in laboratory-scale tanks, the ow characteristics that enhance particle suspension
are not as apparent. The mathematical model also predicts dierent solute mixing behavior
between the laboratory and industrial-scale tanks. At low supercial gas velocities (usg
0.003 m/s) the eect of the increasing dd/dt on the decreasing mixing time is smaller in the last
tanks. Hence, according to the calculations, it should be expected that industrial tanks with
dd/dt = 0.1 have advantages in regard to particle suspension in comparison to tanks with
dd/dt = 0.5 and, at the same time, they should be comparable in respect to solute mixing under
low supercial air velocities, at which they also exhibit good energy eciency.
DOI: 10.1007/s11663-007-9079-5

The Minerals, Metals & Materials Society and ASM International 2007

I.

INTRODUCTION

PACHUCA tanks are hydrometallurgical reactors in

which the suspension of mineral particles and the mass
transfer processes controlling the kinetics of the reactions are intimately linked to the motion of the liquid
that results from the injection of gas through the base
and, in general, into a central draft tube. These tanks are
large cylindrical vessels, commonly 4 to 10 m in diameter and 15 m in height, that normally include a tapered
base with a 30- to 60-deg half angle and a draft tube
typically 0.4 to 1.0 m in diameter.[1,2,3] These reactor
vessels are widely used in the hydrometallurgical industry to carry out gold leaching, uranium leaching, the
bacterial oxidation of pyrite, copper leaching, and
celestite conversion, among other processes. Their
relevance has been claimed to be due to their simple
construction, to the absence of moving parts,[4] and
additionally, according to Weiland,[5] to their low energy
E. RODRIGUEZ M., Postdoctoral Student, A.H. CASTILLEJOS
E., Professor, and F.A. ACOSTA G., Associate Professor, are with
the CINVESTAVUnidad Saltillo, 25000 COAH, Saltillo, Mexico.
Contact e-mail: humberto.castillejos@cinvestav.edu.mx
Manuscript submitted April 4, 2007.
METALLURGICAL AND MATERIALS TRANSACTIONS B

consumption. However, this author[5] and others[1,6]

indicate that if the ratio of the slurry ow rate to the
air ow rate is taken as a measure of the energy
performance of the Pachuca tank, this will be high at
low gas ow rates, but will become poor after a critical
air ow rate is exceeded. In fact, the energy performance
will depend on the particular requirements of the
process. Fraser[6] points out that a typical Pachuca tank,
operating at an air rate of 0.02 Nm3/min/m3 of tank
volume, may not be able to satisfy the necessary oxygen
uptake rate for gold leaching, unless excess air is applied
with the consequent waste of energy. For other processes, such as the acid leaching of uranium ores[4] and
celestite conversion,[7] in which the purpose of the air is
the agitation and suspension of particles and not
aeration, the process goal may be achieved with a
satisfactory energy performance. Lamont[4] indicated
that operations requiring mainly particle suspension and
scrubbing are best carried out in full-center column
Pachuca tanks, while those needing aeration benet from
the use of free-airlift or stub column tanks. Therefore,
based on the needs of a particular leaching process, the
operating and design guidelines have to vary.
In the Pachuca tank, the motion of a slurry of mineral
particles is established when gas is centrally injected

through its base into a draft tube, to produce a liquid

upow within this region and a downow in the annular
space formed between the draft tube and the tank wall.
The driving force for this circulation is the imbalance in
the hydrostatic heads between these two regions.
Clark[1] and Zaisha et al.,[8] working with Pachuca
tanks with a tapered base, found that, at low supercial
gas velocities, the ow pattern of the gas phase in the
draft tube corresponds to the bubble ow regime. But
once the gas velocity increases above a certain limit, the
ow regime changes to churn-turbulent or slug. Clark[1]
pointed out that the change in the gas ow regime aects
the slurry circulation velocity. Weiland[5] has indicated
that liquid circulation and distribution of the gas phase,
as well as the rates of heat, mass, and momentum
transfer in Pachuca tanks, are determined by a complex
interaction between buoyancy, inertia, friction, and
hydrostatic pressure forces; these forces are determined,
ultimately, by the operating and design parameters. In
his experimental work involving a tank with a at
underside, the author concluded that the ratio of the
draft tube to the tank diameters must vary according to
the process requirements. He found that dd/dt ratios
between 0.8 and 0.9 favor mixing and oxygen mass
transfer, while a ratio of 0.6 enhances large liquid
velocities and good particle suspension. Koide et al.[9,10]
reported similar results for this diameters ratio, indicating that the critical gas velocity, dened as the supercial gas velocity required for the complete suspension
of solid particles,[10] has a minimum at a diameters ratio
of 0.6. In contrast to these results, Roy et al.[11] found in
a laboratory study that the critical air supercial
velocity for particle suspension tends to a minimum at
dd/dt ~ 0.1 to 0.2. The authors indicated that this
diameters ratio becomes closer to 0.1 (the ratio prevalent
in industrial practice) as the tank diameter increases.
The critical velocity was found to decrease with the
scale-up (increase in diameter) of the tank, i.e., the tanks
become more energy ecient as their diameter increases,
maintaining a tank height/tank diameter ratio, ht/dt,
constant. It was suggested that this behavior stems from
the increase in the supercial liquid velocity in the draft
tube that takes place with the increase in tank diameter,
for a given air supercial velocity, usg, and for given dd/
dt and ht/dt ratios. In another study concerning oxygen
mass transfer, Roy and Shekhar[12] found that the eect
of the dd/dt ratio (in the range of 0.1 to 0.3) on the
oxygen mass transfer coecient is also aected by the
tank scale.
A study on the gas holdup and liquid circulation rate
in bubble columns concluded that the overall gas
holdup values are smaller and less dependent on the
properties of the liquid or liquid-solid phases in columns
with a draft tube than in columns without one.[13] It has
been observed that the air holdup increases with an
increase in supercial gas velocity,[13,14,15] but decreases
with an increase in the dd/dt and ht/dt ratios.[14] Local
measurements of gas holdup showed that, in the
homogeneous bubble regime, this parameter was uniform over the bubble column cross section, but for other
regimes the holdup proles were symmetrical, with the
maximum located at the center.[13] The mixing of a

solute tracer was studied in laboratory air-lift loop

reactors,[5] and the results showed that the mixing time
decreased as the dd/dt ratio increased, within the range
studied (0.59 to 0.88). Furthermore, the investigator[5]
reported that the mixing time decreases rapidly with the
increase in supercial gas velocity, until this reaches a
value after which any additional decrease becomes
negligible. The scale of the tank also inuences the
eect that the design and operating variables have on
the mixing of solutes. Hence, the eective mixing time,
dened as the mixing time per unit reactor volume,
decreases with an increase in the reactor volume, and
the inuence of the dd/dt ratio on this parameter
becomes also smaller.[5]
Although many experimental investigations[114] have
been carried out about the inuence of design and
operating variables on uid dynamics, mass transfer,
and particle suspension in Pachuca tanks, less frequent
but equally important eorts have gone into predicting
these phenomena with the purpose of advancing the
design of these reactors. Thus, Lamont[4] developed a
macroscopic energy balance to estimate the liquid
circulation ow rate; his analysis established a balance
between the energy input by the air, as it expands while
rising through the vessel, with that consumed in
overcoming the hydraulic heads associated with the air
discharge, draft tube friction, and pulp velocity. In a
latter work, Clark[1] calculated an average slurry velocity establishing a macroscopic momentum balance, over
the draft tube length, by equating the hydrostatic head
over the height of the tank to the sum of the pressure
gradients associated with the frictional wall losses,
entrance losses, and acceleration losses of the slurry in
the draft tube. The author arrived at a design equation
that relates the supercial slurry velocity to the supercial air velocity. In a pioneering work, Shekhar and
Evans[16] calculated the single-phase ow motion in the
annulus by solving the turbulence NavierStokes equation. The two-phase region inside the draft tube was
excluded from their analysis by specifying ad-hoc
boundary conditions at its top and base. The model
aorded a qualitative insight into the problem but
produced computed ow patterns that were signicantly
dierent from those measured in the laboratory.[3] Singh
and Mazumdar[17] modeled the ow problem by solving
the Eulerian motion equation for the liquid, coupled
with the Lagrangian motion equation for the gas
bubbles and the k-e model of turbulence. Their predictions regarding the change in the supercial liquid
velocity in the draft tube with the supercial gas velocity
were reasonable; however, the calculated ow pattern in
the annulus did not resemble the experimental one.[3] On
the other hand, the use of a simple algebraic model for
estimating the eective viscosity aorded a better
representation of the ow pattern in the annulus, but
then failed to predict the average circulating velocity in
the draft tube. The authors[17] concluded that the key
to a proper mathematical representation of the uid
ow in Pachuca tanks would lie in a more precise
modeling of turbulence within the tank annulus. However, previous work[18] has indicated that the modeling
of bubble-driven recirculating ows is more sensitive to
METALLURGICAL AND MATERIALS TRANSACTIONS B

an accurate specication of the spatial distribution of

the gas volume fraction.
The literature review shows that mixing,[5] particle
suspension,[11] and oxygen mass transfer[12] are aected
by design and operating variables to a degree that
depends on the size-scale of the Pachuca tank. Hence,
there is a need for a mathematical model able to
simulate the behaviors observed in the laboratory; that
model can then be used to set design and operating
criteria for industrial reactors. This article presents both
laboratory and computational work on uid ow and
mixing in Pachuca tanks. Apart from encompassing the
turbulent mass transfer problem, the main dierence of
the present mathematical model with respect to previous
ones[16,17] is that the gas volume fraction within the draft
tube was evaluated from the simultaneous solution of
the turbulent NavierStokes equation for a single-phase
ow with variable density and the drift ux model[19] for
gas-liquid two-phase ow. The model predicts fairly
closely the water velocity elds, the evolution of the
solute concentration at the sensor location, and the
mixing times for the dierent experimental conditions
investigated. Also, the calculated average gas holdups
exhibit good agreement with the results from experimental correlations.[20] According to the computational
results, laboratory and industrial-scale tanks exhibit
signicant dierences in the uid ow behavior of the
liquid, so the laboratory results should be carefully used
when trying to understand the functioning of industrial
tanks.

II.

EXPERIMENTAL WORK

A. Experimental Apparatus and Conditions

A schematic illustration of the experimental setup is
shown in Figure 1. The laboratory Pachuca tank was
made of Plexiglas and had a conical base at the apex of
which a nozzle, 0.002 m in diameter, was placed for the
injection of air. For measuring the air ow rate, a
rotameter was employed, which maintained the air
pressure at 103.4 kPa (15 psig) in all the experiments. A
single tank with a diameter of 0.28 m and an underside
with a 45-deg half angle was built with sucient height
to allow the selection of the water bath level with respect
to the base apex, ht. Dierent draft tubes were used to
obtain the desired combination of dd/dt, ht/dt, draft tube
submergences, sd, and draft tube-to-base separations,
sdb. These design parameters are indicated in the gure.
The Pachuca tank was enclosed in a larger rectangular
vessel lled also with distilled water, to minimize
parallax eects that would aect negatively the visualization of images. For studying the mixing of a solute, a
tilting cup was placed slightly above the bath surface, to
pour a tracer solution. The center of the cup was
separated from the tank wall by 0.058 m, and aligned
with it but 0.35 m below the bath surface a conductivity
cell was installed. All this is indicated schematically in
Figure 1.
The uid ow and mixing determinations were done
under the operating and design conditions listed in
METALLURGICAL AND MATERIALS TRANSACTIONS B

Table I. From the table, it is seen that a wide range of

conditions were covered and that the usg, dd/dt, and ht/dt
were similar to those of industrial interest cited by
Shekhar and Evans.[3]
B. Experimental Procedures: Fluid Flow Measurements
As seen in Figure 1, the local liquid velocity was
measured using a particle image velocimetry (PIV)
system based on a diode laser head (Oxford Lasers,
HSI 1000, Oxford Lasers Ltd., Didcot, UK), which
produces an infrared light sheet (0.05 m high and
0.003 m thick) for illumination of the ow; an electronic
unit to control the number, duration, and separation of
laser pulses; a charge-coupled device (CCD) camera
with a resolution of 1035 (V) 1317 (H) pixels (Kodak
Megaplus 1.4, Eastman Kodak Co., Rochester, NY,
USA); a camera-laser interface; a computer with a frame
grabber (Data Translation DT3851, Data Translation
Inc., Marlboro, MA); and software for the acquisition
and analysis of PIV data (AEA Technology, Visiow,
AEA Technology, Didcot, UK).
For the experimental runs, the water in the tank was
seeded with Pliolite (The Goodyear Tire and Rubber
Corp., Akron, Ohio) particles (with a density and
diameter of 1040 kg/m3 and 63 lm, respectively) by
adding ~0.1 kg/m3. Then, the air was turned on to
provide the desired usg and, once steady-state conditions
were achieved, the measurement of the local liquid
velocity was started. The local velocity vectors were
determined from the analysis of double-exposure photographs of tracer particles, which showed them at
sequential positions over the plane formed by the thin
sheet of laser light. A laboratory-made camera-laser
interface was used to synchronize the camera with the
laser and generate laser pulse separations from 3 to
20 ms, which were required to detect the low velocities
found in the system (~0.01 to 0.4 m/s). The duration of
the pulses was 10 ls. Figures 2(a) and (b) show doubleexposure photographs of seed particles passing within
the draft tube. When bubbles were present in the eld of
view, the seed images appeared blurred and the light was
scattered by the bubble surface, as seen in Figure 2(a).
In contrast, when bubbles were absent (Figure 2(b)),
sharp seed image pairs were obtained, from which the
local mean water velocities could be determined. This
was done using the autocorrelation method implemented in the PIV data analysis software;[21] in carrying
out this analysis, every frame (0.020 (V) 0.027 (H) m2)
was divided into 21 (V) 28 (H) interrogation regions.
Thus, the analysis of every frame produced a map of 588
velocity vectors. The validation of the vectors was done
by comparing each vector with those close to it; a vector
was accepted if its magnitude and direction fell within
the bounds dened by visual inspection of the vectors in
the respective frame. Spurious vectors were replaced by
estimates obtained by interpolation of neighboring ones.
The directional ambiguity associated with the autocorrelation analysis of double-exposure images was solved
by visual inspection of the ow using a high-speed
camera with motion analysis capabilities (Kodak
Ektapro, Eastman Kodak Co., Rochester, NY).

Fig. 1Schematic diagram of experimental setup.

Table I.

Operating and Design Conditions and System Properties Used in the Experimental and Computational Work

usg 103 m/s

Experimental 1, 4, 7
Computational 0.125, 0.5, 1, 4, 7
Experimental and computational
1, 1.7, 2.4, 3.8, 5, 7
0.125, 0.5, 1, 4, 7
qg kg/m3
ql kg/m3
l kg/m s
rC

dd/dt

ht/dt

Liquid uid ow eld determination

0.25, 0.5
3, 3.14, 3.23, 3.43
Solute mixing evolution and mixing time determination
0.1, 0.25, 0.5
2, 3
Simulated industrial conditions
0.1, 0.25, 0.5
2, 3
Fluid properties
1.07
998
1 10-3
521.65

C. Experimental Procedures: Mixing Time

Measurements
In these experiments, 3 mL of a saturated KCl
solution were added after stable uid ow conditions
were established. The change in the electrical conductivity of the bath caused by the addition of the tracer
solution was measured by an electrical conductivity
probe and recorded in a computer in the form of voltagevs-time traces. The probe consisted of two stainless steel
wires, 0.002 m in diameter and 0.02 m in length, embedded in an epoxic resin and separated by 0.004 m.[22] The
output of the conductivity meter was in the 0- to 1-V
range and was proportional to the solute concentration.
The current frequencies at which the probe was operated
were maintained above 1 kHz, to avoid polarization that
would cause wrong measurements.[23]

sd (m)

sdb (m)

0.07, 0.15

0.025, 0.05

0.07

0.025

0.94

2.4

The mixing time was dened as the time required to

attain a 95 pct level of homogeneity, which was specied
in terms of the degree of mixing, Y, as





V1
Vt



1
Y 1
Is 1



V1
Vo

where the intensity of segregation, Is, was expressed in
terms of voltage values, because of the proportionality
of these values to concentration.
III.

MATHEMATICAL MODEL

A. System Considered and Assumptions

Before performing calculations, ow visualizations
were carried out during the injection of air in the
METALLURGICAL AND MATERIALS TRANSACTIONS B

Fig. 2Double-exposure photographs of seed particles moving

through the draft tube at (a) a moment at which bubbles were
present in the eld of view and (b) a moment at which bubbles were
absent from the eld of view.

Pachuca tanks, with draft tube diameters of 0.25 dd and

0.5 dd, as shown in Figures 3(a) through (d). From the
photographs, it can be seen that the air-water ow
evolves from the bubbly regime to the churn-turbulent
regime as the gas ow rate increases, with the bubbles
tending to occupy the whole draft tube diameter as this
decreases, i.e., the ow turns to the slug regime. In these
systems, the bubbles rise as a column in a fashion very
dierent from the laterally-spreading plumes that are
observed in physical models of gas-stirred ladles.[24]
Also, although the gas volume fraction was not measured quantitatively, video lms of the rising bubbles
suggested that, on a time-averaged basis, there was little
variation in the gas volume fraction across the diameter
and height of the draft tube; only at the higher ows did
the bubbles accumulate above the tube to form a bubble
layer with a higher gas holdup. Other features that
should be noticed from the photographs are that, except
for the highest ows (and, in these cases, only at the
top), the annular region was free of bubbles and the free
surface remained relatively at. The physical properties
of the system are given in Table I.
Based on these observations, it was assumed that, to
establish a uid ow model, the motion of the water is
essentially buoyancy-dominated and the bubble column
METALLURGICAL AND MATERIALS TRANSACTIONS B

Fig. 3Photographs illustrating the air-water ow regimes appearing in Pachuca tanks operating under dierent gas ow rate conditions and with draft tubes of dierent diameter. Bubbly regime at
Qg = 0.62 10-4 m3/s (usg = 1 10-3 m/s) for (a) dd = 0.07 m and
(b) dd = 0.14 m. Churn-turbulent regime at Qg = 4.31 10-4 m3/s
(us g = 7 10-3 m/s) for (c) dd = 0.07 m and (d) dd = 0.14 m.

is axisymmetric and develops in a short time, so that the

conditions of two dimensions (2-D) and steady state
were considered; the integration domain extended over
the axisymmetric plane of the tank. Furthermore,
Pachuca tanks have a well-dened boundary between
the one-phase and the two-phase region, and the last one
has a uniform gas volume fraction distribution across
the radial direction, according to the results of Gavroy
et al.[13] and the ow visualizations reported earlier.
Therefore, it was assumed that the gas holdup in the
draft tube varies only in the axial direction according to
the drift ux model introduced by Deb Roy et al.[19] for
gas-stirred ladle systems, in the form



a

1
2p Rrd

Qg

Table II. Meaning of Dependent Variables / and

Expressions for C/ and S/ in the Governing Equations
for Turbulent Flow

2

ru Uslip dr

Hence, the ow domain was assumed to be occupied

by a single uid of variable density, q, specied as
aqg ; for
q 1


aql

0  r<rd

q ql ; for rd tw <r<rt

3

According to the literature,

the slip velocity,
Uslip, of bubbles in the slug, churn-turbulent, and
bubbly regimes span in the range of 0.18 to 0.45 m/s.
In this work, a value of 0.40 m/s was assumed, similar to
studies on gas-stirred ladle systems.[19,27] Also, since the
bath-free surface remained relatively quiescent over the
majority of the gas ow rates of interest, it was assumed
that this was at and exerted a negligible shear on the air
phase above it.
Solute mixing is a time-dependent problem and, in the
specic case studied here, was in three dimensions (3-D),
due to the position at which the solute was added.
Hence, the integration domain was extended to the
whole tank described in r-h-z coordinates. In the
circumferential direction, solute mass transfer was
supposed to occur only by turbulent diusion. The
eective diusivity, De, was assumed to be the same
along the three directions and was evaluated from the
solution of the 2-D turbulence uid ow problem as
given in Section IIIB. Also, since the solute, KCl,
dissolves only in the liquid phase, the whole domain was
assumed to be occupied by liquid, so that q = ql,
everywhere. The gas phase was taken into account by
blocking the fraction of the discretization cells occupied
by it, such that the local liquid volume was given as

leff
leff

k
e

l
l

Deff

eff
eff

Continuity equation
0
NavierStokes equation



@
@v
v
@
@u

@p
@r r@r rleff @r
2leff r2 @z leff @r

 @

@
@v
@u

@p
a
@z r@r rleff @z @z leff @z ql g

k-e turbulence equations
/rk
G-qe
2
/re
C1 ke G
C2 q ek
Solute conservation equation
0

The turbulent viscosity lt was evaluated by the ke

model[28] in its conventional form, using the constants[29]
displayed in Table III. Hence, the value of lt at any
point in the eld was calculated from the local values of
k and e according to
lt

Cd qk2
e

8

The eective transport coecient of the solute was

given as
l
l
CC;eff
t
9
rC rC;t
where the turbulent Schmidt number for concentration
rC,t ~ 1.
C. Initial and Boundary Conditions
1. Fluid ow problem
At the axis of symmetry, zero ux conditions were
imposed, as
v

@u @k @e

0
@r @r @r

10

At the bath-free surface, the axial velocities and the

stresses were assumed equal to zero, as in

Under the assumptions stated here, the governing

dierential equations for continuity, momentum balance, solute balance, and the turbulence quantities can
be written in the general form
6

@v
@k @e
u

0
@z
@z @z

11

At the walls and base, impermeable and nonslip

conditions were imposed, as in
vuke0

where / stands for the dependent variables given in

Table II, together with the expressions for the transport coecients and source terms. The transient term
in the previous expression is zero for all the dependent
variables /, with the exception of the solute concentration C. The eective viscosity appearing in the table is
given by
leff l lt

v
u

S/

5

B. Governing Equations

@q/
r  qU/ r  C/ r/ S/
@t

4

[1,25,26]

a
Vl V1

C/

7

12

To treat the steep variation in the momentum and

scalar transport properties that occur close to solid
Table III. Constants in the k - e Model
C1
1.44

C2

Cd

rj

re

1.92

0.09

1.0

1.3

METALLURGICAL AND MATERIALS TRANSACTIONS B

walls, the near wall treatment suggested by Launder and

Spalding[29] was applied.
2. Mixing problem
The initial condition established that the tracer was
added as a pulse so that, at time t = 0, a number of
discretization cells with a volume equal to the addition
volume (i.e., 3 mL and 76 L for the experimental and
industrial-scale cases, respectively) acquired the concentration of the tracer solution; these cells were located
just below the addition cup. In the rest of the integration
domain, the initial solute concentration was zero. The
boundary conditions specied that all the bounding
surfaces were impervious to the solute tracer. Furthermore, since the integration domain covered the whole
Pachuca tank, cyclic boundary conditions were imposed
at two r-z planes, B1 and B2, separated by 360 deg, so
that
CB1 CB2

13

D. Numerical Procedure
The governing equations for uid ow and solute
mixing were solved uncoupled and in axisymmetrical
body-tted coordinates by the control volume method
implemented in the PHOENICS code (CHAM, Ltd.,
Wimbledon, London, UK).[30] Depending on the values
of the design parameter, the discretized grids for the
experimental tanks varied between 56 77 through
56 92 and 56 77 47 through 56 107 47 nodes
for the ow and mixing problems, respectively; for the
industrial tanks, however, the meshes were 168 80
through 100 134 and 168 80 125 through
100 134 125. Time steps of 0.5 and 1.0 seconds
were used for solving the mixing problem in laboratory
and industrial-scale, respectively. These selections ensured grid-independent results in the dierent problems
solved.
The convergence criterion over the domain was set as
X




E/P
 10
3
14
Domain

where E/p is the residual of the dependent variable /

at position P.

IV.

RESULTS AND DISCUSSION

In this section, we present experimental and computational results to discuss the eect of operating and
design variables on several parameters characterizing
uid ow and mixing in Pachuca tanks; simultaneously,
both types of results are compared. In the case of the
velocity eld, the evolution of the solute concentration
at the measuring point, and the gas holdup, only a
selection of results is presented. However, in the case of
the liquid circulation velocity and mixing time, the
outcomes of all the conditions studied are given. Finally,
predictions for industrial-scale tanks are considered.
METALLURGICAL AND MATERIALS TRANSACTIONS B

A. Time-Averaged Velocity Field and Volume-Averaged

Gas Holdup
Figures 4(a) and (b) present a comparison between the
theoretically-predicted and experimentally-measured
velocity elds, for two conditions involving dierent dd/
dt ratios and gas supercial velocities. It is seen that the
agreement between the measurements and the predictions is good, both in the magnitude of the local velocities
and in the ow patterns of the water phase. From the
gures, it can be noticed that the recirculating loop that
forms on the upper part of the annular region extends
over a larger length for tanks with a small dd/dt ratio.
Additionally, it is observed that the experimental velocity
elds display a somewhat disorderly liquid motion in the
draft tube that can be attributed to large-scale turbulence
uctuations caused by the pass of bubbles. Moreover, the
bubbles during their rise have a swaying motion as a
result of interaction with other bubbles, which causes
them to deform, break, and coalesce, and thus move the
liquid in directions dierent from the main direction.
A stricter comparison between the experimental and
computed velocities is given in Figures 5 and 6, where the
axial velocity components are plotted as a function of the
radial position for several heights along the tank and for
the conditions corresponding to the previous gure. It is
seen that, in the annular region, the accord between
the predictions and the measurements is excellent. The
discrepancies occur mainly in the draft tube at the
highest gas supercial velocities; nevertheless, they are
relatively small, despite the simple representation of the
two-phase region. From the gure, it is also observed
that, except for the lower part of the draft tube, the
measured and computed proles of the axial velocity are
relatively at, with the liquid moving in plug ow.
As was mentioned previously, one of the assumptions
of the uid ow model was that the slip velocity of the
bubbles was equal to 0.40 m/s, regardless of the airliquid ow regime prevailing at each particular set of
conditions. Therefore, it is appropriate to notice here
that the computational experiments showed that halving
this velocity increased the predicted liquid velocities in
the draft tube by just 10 to 15 pct, i.e., the assumption
did not strongly aect the computational results.[31] To
explore further the adequacy of the drift ux model for
representing the two-phase region, the predicted volumeaveraged gas holdup within the draft tube, dened as
R hd

adz

avol 0
15
hd
was compared with the experimental results reported
by Shekhar[15] and with the values calculated from a
correlation derived by Hills.[20] The correlation obtained for bubble columns was rewritten[15] for Pachuca tanks in the form
usg;d

avol
16
1:35usg;d usl;d 0:93 0:24
where the liquid supercial velocity usl,d was evaluated
as the volume average of the predicted uz velocity component over the draft tube, as

Fig. 4Predicted and measured mean velocity elds in experimental Pachuca tanks under the following conditions: (a) usg = 0.001 ms-1, dd/
dt = 0.25, ht/dt = 3.0 and (b) usg = 0.007 ms-1, dd/dt = 0.5, ht/dt = 3.14. The others parameters are sd = 0.07 m and sdb = 0.025 m.

Fig. 5Comparison between experimental measurements () and numerical predictions (-) of the radial variation of the axial velocity component
of the water phase, at various axial positions.

METALLURGICAL AND MATERIALS TRANSACTIONS B

Fig. 6Comparison between experimental measurements () and numerical predictions (-) of the radial variation of the axial velocity component
of the water phase, at various axial positions.

R rd R hd

uz 2prdrdz
usl;d R0 r R0 h
d
d
0 0 2prdrdz

17

The results are shown in Figure 7 for two sets of

conditions. From the graph, it is seen that the avol values
predicted by the model agree well with those obtained
by the correlation, and both show that the gas holdup
decreases with the increases in the ht/dt, ratio, in
agreement with the literature.[14] This behavior stems
from the increase in the buoyancy and expansion power
of the bubbles as the bath depth increases and causes an
increase in the liquid upow in the draft tube and,
hence, a faster travel of the bubbles through the bath.
The gure shows that the comparison of the model with
the experimental results[15] is quite good for an ht/
dt = 2.1, but not for a 1.4 ratio. However, in this last
case, the measured values did not comply with the trend
expected from the eect of the ht/dt ratio on the gas
volume fraction.
B. Effect of dd/dt and ht/dt on Water Flow Circulation,
Supercial Velocities, and Gas Holdup
Figure 8 shows a plot of the circulating water ow
rate, Ql, as a function of the injected gas ow rate, for
two dd/dt ratios and several ht/dt ratios. The graph
includes experimental and predicted values, and it is
seen that the agreement between both types of results is
METALLURGICAL AND MATERIALS TRANSACTIONS B

Fig. 7Plot of the variation of the volume-averaged air holdup in

the draft tube as a function of supercial air velocity, comparing
experimental, correlation, and model results. The conditions used involved a Pachuca tank with a base cone angle of 45 deg.

excellent. For the range of the parameters ht/dt and sd

shown in the gure, it is seen that their eects on liquid
circulation are unimportant, but as it is seen later in this
article, the inuence of ht/dt is appreciable when it varies
over a wider range. On the other hand, the gure shows
that the diameters ratio dd/dt has a very signicant eect

Fig. 8Circulating water ow rate as a function of air ow rate, for

experimental Pachuca tanks with dierent design parameters.

on water circulation. This result indicates that, with the

increase in dd/dt ratio, the pressure losses in the draft
tube decrease, so that large circulation ows are
obtained despite the decrease in the pressure imbalance,
a, responsible for the ow. This argument arises from
ql g

the calculated variation of the area averaged gas
fraction,

a, as a function of the z-position, which is
presented in Figure 9 for several dd/dt ratios. From the
gure, it is seen that the model predicts that

a decreases
with the increase in dd/dt, in agreement both with the
experimental results of Roy et al.[14] and with the visual
observations shown in the photographs of Figure 3.
Additionally, it should be noticed from Figure 9 that the
gas holdup values predicted by the present model are

very close to the a
vol values calculated from Eq. [16],

which are displayed in the key of the gure.
The eect that the distance between the lower end of
the draft tube and the base of the tank, sdb, has on the
water circulating ow was also simulated appropriately
by the model, as it can be seen from Figure 8. For the
situations modeled, it is observed that the resistance to
the ow imposed by the passage formed between the
draft tube and the tank base has a signicant eect on
water circulation. As the water moves from the straight
section of the annulus to the tapered one, it converges in
the lower end of the draft tube, changing its direction
and accelerating. Once the water crosses the passage it
diverges and changes its direction drastically, as can be
appreciated from the velocity elds presented in Figure 4. Therefore, the model results indicate that the
increase in sdb tends to moderate the ow changes and
reduce the energy losses, resulting in an increased water
circulation ow, as can be appreciated from Figure 8.
This predicted behavior agrees with the trends of the
experimental ndings of Koide et al.[9]
The importance of knowing the circulating ow rate
of the water in Pachuca tanks is twofold: (1) the
circulating ow rate constitutes an index of the rate of
transport of the chemical species by bulk ow and (2) it
would have a direct inuence on the dynamics of the
dispersion of the solid particles by determining the
supercial water velocities in the annulus, usl,a, and in
the draft tube, usl,d. From the mass conservation balance
of water, Ql can be written as
Ql Aa usl;a Ad usl;d

18

Large usl,d values would produce a large upward drag

of the particles in the draft tube, while a low usl,a
would cause a small downward drag of the particles in
the annular region; this combination of factors would
promote lifting and a reduced sedimentation of the
particles in the draft tube and annular section, respectively. Hence, from Figure 10, which shows the variation in usl,a and usl,d with the supercial air velocity, it
may be expected that small dd/dt (~0.1) ratios would
favor better particle suspension than would large ones.
Furthermore, the gure shows that the absolute values
of the supercial velocities and, as a consequence, the
water circulation ow rate increase with the ht/dt, for a
given supercial air velocity. A qualitative insight
about the eect that an increment in this parameter
would have on particle dispersion is dicult to assess
from the information provided by the gure, since this
suggests that it would simultaneously enhance particle
lifting in the draft tube and sedimentation in the annulus. However, this and other aspects of particle dispersion are currently under investigation.[31]

C. Effects of usg, dd/dt, and ht/dt on Solute Mixing

and Flow Turbulence
Fig. 9Variation in gas volume fraction with axial position for Pachuca tanks with dierent dd/dt and ht/dt ratios and for sd = 0.07 m,
sdb = 0.025 m, and Qg = 0.616 10-4 Nm3/s (usg = 1 10-3 m/s).

From the results just presented, it should be noticed

that the numerical solution of the turbulent Navier
Stokes equation provides a great deal of information on
METALLURGICAL AND MATERIALS TRANSACTIONS B

Fig. 10Predicted supercial water velocities vs supercial air velocity, comparing the eects of dierent dd/dt and ht/dt ratios.

the velocity elds and the gas holdup distribution, and

suggests how the bulk liquid motion may aect solute
mixing and particle suspension. On the other hand, this
section discloses how the solution of the solute mass
conservation equation complements that information by
distinguishing the contributions of the bulk liquid
motion and the turbulence on solute mixing and by
pointing out the eect that the operating and design
parameters have on it.
Figures 11(a) and (b) show comparisons between
experimentally-measured and theoretically-predicted tracer dispersion curves for laboratory Pachuca tanks with
ht/dt ratios equal to 2 and 3, respectively. It may be seen
that, while the mathematical model does not exactly
match the measured curve, it does predict the overall
trend quite well; in many cases, it is quite accurate in
predicting the time required for the complete homogenization of the aqueous solution, as shown in Figures 12(a)
and (b). In these gures, it is seen that, with the increase in
supercial air velocity, the mixing improves rapidly until
the velocity is about 3.8 10-3 m/s, while at higher
velocities, mixing becomes less ecient.
Comparing Figures 12(a) and (b), it is appreciated
that the mixing time increases with increases in the tank
size, i.e., ht/dt, despite the supercial liquid velocities
augments (i.e., the water circulation ow rate) with the
increase in this ratio, as reported in Figures 10(a) and
(b). The turbulence energy elds shown in Figure 13
suggest that this behavior arises because the zone with
low turbulence levels (found in the lower part of
METALLURGICAL AND MATERIALS TRANSACTIONS B

Fig. 11Evolution of solute tracer dispersion at the conductivity

sensor position for Pachuca tanks with (a) ht/dt = 2 and (b) ht/dt =
3; other conditions are the same.

annulus) occupies a greater proportion of the tank as

ht/dt increases, for a given usg and dd/dt ratio. Hence, as
expected, turbulence plays a very important role in the
complete mixing of solutes.
Additionally, the measured and predicted results
given in Figures 12(a) and (b) show that, for a given
supercial air velocity, the inuence of the diameters
ratio dd/dt on the time of mixing decreases with the tank
size, i.e., the mixing time decreases as the dd/dt increases
in tanks with ht/dt = 2, while it is almost the same for
diameters ratios equal to 0.25 and 0.5 in tanks with ht/
dt = 3. Apparently, the reason for this behavior is,
again, the variation in the turbulence level and the
turbulence pattern with the tank conguration. As the
plots of Figure 13 show for the smaller dd/dt ratio, the
turbulence level at the base of the draft tube is larger and
the regions of higher turbulence at the top of the
annulus are more extensive compared to those for the
bigger ratio; additionally, these changes are more acute
for the larger ht/dt ratio. This behavior indicates that the
relative importance of water circulation velocity and
turbulence on mixing time is a function of tank size and
geometry.
D. The Predicted Behavior of Industrial-Scale Tanks
Since the previous sections show that the uid ow
and mass transfer models give a good account of the
behaviors observed in the experiments, this section

Fig. 12Measured and predicted mixing time as a function of

supercial air velocity for laboratory Pachuca tanks with dierent
diameters ratios and for (a) ht/dt = 2 and (b) ht/dt = 3.

presents the predictions for industrial-scale Pachuca

tanks, the ultimate goal of this work.
The calculations for the industrial-scale tanks have
been done for three supercial air velocities and for a
particular set of tank dimensions that is within the range
used in the industry. The computations involved two
diameters ratios; for comparison with laboratory-scale
tanks, the areas ratios Act/Aa and Acb/Aa in both scales
were the same, so that energy losses due to the ow
reversal at the upper and lower ends of the draft tube
should be expected to be similar.
Figures 14 and 15 show the velocity elds computed
for the experimental and industrial-scale tanks for dd/dt
equal to 0.1 and 0.5, respectively. For the small
diameters ratio (Figure 14), it is seen that the calculated
fraction of the tank height occupied by the recirculating
loop formed in the annulus is much larger for the
industrial-scale tank than for the equivalent laboratoryscale tank. The large velocities at which the liquid exit
the draft tube cause strong currents to hit the tank wall,
which then move backward against the draft tube to
recirculate. From the large recirculating loop formed in

the industrial-scale tank, it should be expected that, if

particles were present, a large proportion of them would
be maintained in suspension, while those moving to the
base of the tank through the annulus would be lifted by
the large ascending velocities once they reach the
entrance of the draft tube. For dd/dt = 0.5, the predicted ow patterns for tanks of both scales are very
similar, with the height fraction of the recirculating loop
being relatively short, as can be appreciated from
Figure 15. In this case, it could be thought that, if
particles were present, a large proportion of them would
move mainly downward, settling on the tank underside
and possibly hindering the entrance into the draft, where
they would be picked up by the ascending water
currents. Therefore, it should be expected that a dd/
dt = 0.1 would be more suitable for particle suspension
than would be a larger ratio.
Figure 16 shows the ratio of the supercial liquid
velocity in the annulus to the supercial gas velocity, as
a function of the latter. From the gure, it is seen that
the model predicts that, in industrial-scale tanks, the
water circulation velocity that can be achieved at a given
supercial air velocity is larger than that obtained in the
laboratory-scale tanks. The predictions show that the
eciency of the industrial tanks with respect to their
equivalent laboratory tanks is larger for dd/dt = 0.1
than it is for tanks with dd/dt = 0.5, at low supercial
air velocities (0.004 m/s), while the opposite is true for
larger velocities. Thus, the model indicates that the eect
of the design parameters on the uid ow depends on
the operating conditions and the size scale. Therefore,
the laboratory work should be used cautiously in trying
to interpret the behavior of industrial tanks.
The variation in the mixing time with the supercial
air velocity for dierent dd/dt ratios is shown in
Figure 17 for an industrial-scale tank. Again, the
behavior predicted for large-sized reactors is dierent
from those of laboratory-scale reactors; this was displayed in the plots appearing in Figure 12. In large-scale
reactors and in the range of 0.001 usg 0.004 m/s, the
eect of the dd/dt increases with the increase in the
supercial air velocity, starting from being nil at
usg = 0.001 m/s. From these results, it should be
expected that, at low supercial gas velocities (usg
0.003 m/s), a tank with a dd/dt = 0.1 should present
advantages in regard to particle suspension and energy
eciency in comparison to a tank with dd/dt = 0.5, and
at the same time, should be competitive with respect to
solute mixing. Although according to Figure 16, industrial-scale tanks with a dd/dt = 0.5 involve larger water
circulation velocities and, hence, should be more eective for solute transport by bulk motion, they are not as
eective for providing complete mixing, because of their
smaller turbulence levels, as is seen by comparing
Figure 18(b) with Figure 18(a). From Figure 18(a),
corresponding to dd/dt = 0.1, it may be expected that
the large turbulence levels present at the base of the
draft tube and at the top of the bath should contribute
to homogenize the solute in the bath, particularly by
favoring its transport along the circumferential direction
by turbulent diusion.

METALLURGICAL AND MATERIALS TRANSACTIONS B

Fig. 13Predicted distributions of turbulent kinetic energy in laboratory Pachuca tanks with two dierent ht/dt and dd/dt ratios and a single
supercial air velocity.

V.

SUMMARY AND CONCLUSIONS

This article reports laboratory and computational

work carried out to determine the eect of operating
and design parameters on the motion of the liquid and
the mixing of a solute in Pachuca tanks. The turbulent
recirculating ow model, which uses the drift ux model
to represent the two-phase air-water region, gives a good
representation of the laboratory measurements in regard
to the inuence that operating (usg) and design (dd/dt, ht/
dt, sd, and sdb) variables have on the liquid circulation
velocity, liquid ow pattern, gas holdup, and mixing
evolution and time. The good representation of the
experimental results regarding solute mixing seems to
substantiate the validity of the predicted turbulence
quantities k and e.
The use of the model to simulate the behavior of
industrial-scale tanks revealed that the recirculating
loop that forms in the annular section of the tanks is
more intense and extends over a larger proportion of
METALLURGICAL AND MATERIALS TRANSACTIONS B

the reactor height as the draft tube/tank diameters

ratio, dd/dt, decreases, at the same time the supercial
liquid velocity in the draft tube increases. From the
large recirculating loop formed in industrial-scale tanks
with dd/dt = 0.1, it should be expected that, if particles
were present, a large proportion of them would be
maintained in suspension, while those moving to the
base of the tank through the annulus would be lifted
very energetically by the strong ascending water current, once reaching the entrance of the draft tube.
Moreover, the eciency of industrial-scale tanks in
generating bulk water motion (i.e., usl,a/usg) is greater
than that of laboratory-scale ones. At low supercial
air velocities (0.001 usg 0.004 m/s), the proportion
in which their eciency augments relative to laboratory-scale tanks is larger for dd/dt = 0.1 than for dd/
dt = 0.5. Also, the model indicates that, in industrial
tanks at low usg values (0.001 to 0.003 m/s), the
increase in mixing time with the decrease in the dd/dt
is relatively minor.

Fig. 14Predicted mean velocity elds for (a) experimental tank with dt = 0.28 m, sd = 0.027 m, and sdb = 0.076 m and (b) industrial tank
with dt = 9.60 m, sd = 0.94 m, and sdb = 2.39 m. The other conditions are: usg = 0.001 m/s, ht/dt = 2.08, Act/Aa = 0.040, Acb/Aa = 0.304 and
dd/dt = 0.1.

Fig. 15Predicted mean velocity elds for (a) experimental tank with dt = 0.28 m, sd = 0.027 m, and sdb = 0.076 m and (b) industrial tank
with dt = 9.60 m, sd = 0.94 m, and sdb = 2.39 m. The other conditions are: usg = 0.001 m/s, ht/dt = 2.08, Act/Aa = 0.261, Acb/Aa = 0.70, and
dd/dt = 0.5.
METALLURGICAL AND MATERIALS TRANSACTIONS B

Fig. 16Plot of the proportion of the supercial water velocity in

the annulus with respect to the supercial air velocity as a function
of the latter, for laboratory and industrial-scale tanks.

Fig. 18Predicted distributions of turbulent kinetic energy in industrial-scale Pachuca tanks with (a) dd/dt = 0.1 and (b) dd/dt = 0.5.
The other conditions are: usg = 0.001 m/s, dt = 9.6 m, sd = 0.94 m,
sdb = 2.39 m, and ht/dt = 2.08.

ACKNOWLEDGMENTS
The authors gratefully acknowledge research funding
from the National Council for Science and Technology
of Mexico through Research Grant No. 4407-A9406.
One of the authors (ERM) thanks CONACYT for
receipt of a postdoctoral scholarship. The authors
appreciate the assistance of E.N. Aguilera G. and D.A.
Salinas G. with the laboratory work.

NOMENCLATURE

Fig. 17Predicted mixing time as a function of supercial air velocity for industrial Pachuca tanks with dierent diameters ratios.

Aa, Ad
Act, Acb
Hence, according to our computer model, it should be
expected that industrial tanks with dd/dt = 0.1 would
have advantages in regard to particle suspension in
comparison to tanks with dd/dt = 0.5 and that, at the
same time, they should be comparable in respect to
solute mixing under appropriate operating conditions,
at which they are also energy ecient. Finally, since the
ow characteristics that are responsible for the predicted
behavior of industrial tanks (with dd/dt = 0.1) are not
as noticeable in laboratory-scale tanks, it was very
important to arrive at a mathematical model that allows
quantitative predictions and that can be extended to the
study of particle suspension.

METALLURGICAL AND MATERIALS TRANSACTIONS B

B1, B2
C
C1, C2, Cd
dd, dt, dab
dat
De
g

cross-sectional area of the annulus;

cross-sectional area of draft tube
area of the cylindrical surface of height
sd on top of the draft tube; area of the
cylindrical surface of height sdb below
the base of the draft tube, m2
cyclic r-z planes separated by 360 deg
concentration of solute tracer, kg m-3
constants in the turbulence model
draft tube diameter; tank tube diameter;
thickness of annular section at draft
tube base
thickness of annular section at draft
tube top, m
eective diusion coecient, m2 s-1
acceleration due to gravity, ms-2

G
hd
ht, hd
Is
k
p
Q g, Q l
r, rd, rt
E/P
sd, sdb
S/
t
tw
u
Uslip
usg, usg,d
usl,d
v
V, Vl
Vo, V, V(t)

Y
z

a,

avol
e
/


l, ls, le
q, ql, qg

rate of generation of k per unit volume,

kg m-1 s-3
height of draft tube, m
depth of water bath; height of draft
tube, m
intensity of segregation
turbulence kinetic energy, m2 s-2
pressure, kg m-1 s-2
volumetric gas ow rate at orice
conditions; circulation liquid ow rate,
m3 s-1
radial coordinate; radius of draft tube;
radius of tank, m
residual of the discretized equation for
the ow variable / at node P
draft tube submergence; draft tube to
base separation, m
source term of ow variable /
time, s
draft tube wall thickness, m
mean axial velocity component, ms-1
terminal rising velocity of a single
bubble, m/s
supercial gas velocity with respect to
the tank diameter; with respect to draft
tube diameter, supercial
liquid velocity in draft tube, ms-1
mean radial velocity component, ms-1
volume of discretization cell; volume of
cell occupied by liquid, m3
initial voltage across the electrode of the
conductivity meter corresponding to an
initial solute concentration; voltage
corresponding to complete tracer
dissolution after a suciently long time;
voltage corresponding to a certain local
tracer concentration at time t, V
degree of mixing
axial coordinate, m
cross-sectional area averaged gas
volume fraction; volume-averaged gas
fraction in draft tube
dissipation rate of turbulence kinetic
energy, m2 s-3
ow variable
gradient operator in terms of r-z
coordinates for the continuity,
momentum and turbulence equations
and in terms of r-h-z coordinates for the
mass solute tracer concentration
equation, m-1
divergence operator in terms of r-z
coordinates for the continuity,
momentum and turbulence equations
and in terms of r-h-z coordinates for the
mass solute tracer concentration
equation, m-1
molecular, turbulent and eective
viscosity, kg m-1 s-1
local density; liquid density; gas density,
kg m-3

rC, rC,t
rk, re
h
C/

laminar and turbulent Schmidt numbers

for concentration
Schmidt numbers for k and e
circumferential coordinate, radians
eective exchange coecient of the
quantity /, kg m-1 s-1
REFERENCES

1. N.N. Clark: Miner. Metall. Process., 1984, Nov. pp. 22632.

2. C.J. Hallet, A.J. Monhemius, D.G.C. Robertson: Extraction
Metallurgy 1981, I.M.M., London, 1981, pp. 30820.
3. R. Shekhar and J.W. Evans: Metall. Trans. B, 1989, vol. 20B,
pp. 78191.
4. A.G.W. Lamont: Can. J. Chem. Eng.,1958, Aug., pp. 15360.
5. P. Weiland: Ger. Chem. Eng., 1984, vol. 7, pp. 37485.
6. G.M. Fraser: Aus. I.M.M. Perth and Kalgoorlic Branches, Regional
Conf. on Gold-Mining, Metallurgy and Geology, Oct. 1984,
pp. 24556.
7. F.R. Carrillo P., A. Uribe S., and A.H. Castillejos E.: Miner. Eng.,
1995, vol. 8, pp. 405509.
8. M. Zaisha, Y. Shouzhi, and C. Chen: Proc. 3rd Int. Symp. on
Hydrometallurgy, Research Development and Plant Practice, K.
Osseo-Asare and J.D. Miller, eds., AIME, Atlanta, GA, 1983,
pp. 789806.
9. K. Koide, S. Iwamoto, Y. Takasaka, S. Matsuura, E. Takahashi,
M. Kimura, and H. Hubota: J. Chem. Eng. Jpn., 1984, vol. 17,
pp. 61118.
10. K. Koide, K. Horibe, H. Kawabata, and S. Ito: J. Chem. Eng.
Jpn., 1984, vol. 17, pp. 36974.
11. G.G. Roy, R. Shekhar, and S.P. Mehrotra: Metall. Mater. Trans.
B., 1998, vol. 29B, pp. 33949.
12. G.G. Roy and R. Shekhar: Trans. Inst. Min. Metall., 1996,
vol. 105, pp. C9C15.
13. D. Gavroy, C. Joly-Vuillemin, G. Cordier, and H. Delmas: Trans.
I. Chem. E., 1995, vol. 73, pp. 63742.
14. G.G. Roy, A. Bera, and J.H. Mankar: Trans. Inst. Min. Metall.,
Sect. C, 2000, vol. 109, pp. 9096.
15. R. Shekhar: Ph.D. Thesis, University of California, Berkeley, CA,
1988.
16. R. Shekhar and J.W. Evans: Metall. Trans. B, 1990, vol. 21B,
pp. 191203.
17. H. Singh and D. Mazumdar: Metall. Mater. Trans. B, 1997,
vol. 28B, pp. 72731.
18. J.S. Woo, J. Szekely, A.H. Castillejos, and J.K. Brimacombe:
Metall. Trans. B, 1990, vol. 21B, pp. 26977.
19. T. DebRoy, A.K. Majumdar, and D.B. Spalding: Appl. Math.
Model., 1978, vol. 2, pp. 14650.
20. J.H. Hills: Chem. Eng. J., 1976, vol. 12, pp. 8999.
21. VISIFLOWSystem User Manual, AEA Technology, Oxford,
United Kingdom, 1987, pp. 9:29:5.
22. E.N. Aguilera G: M.Sc., CINVESTAV, Unidad Saltillo, Mexico,
1994.
23. G.G. Murthy and J.F. Elliott: ISIJ, 1992, vol. 32, pp. 19095.
24. A.H. Castillejos E. and J.K. Brimacombe: Metall. Trans. B, 1987,
vol. 18B, pp. 65971.
25. G.F. Hewitt: Measurement of Two Phase Flow Parameters,
Academic Press Inc, London, 1978, pp. 1113.
26. R. Clift, J.R. Grace, and M.E. Weber: Bubbles, Drops and
Particles, Academic Press Inc., NY, 1978, pp. 17173.
27. N. El-Kaddah and J. Szekely: Ironmaking and Steelmaking, 1981,
vol. 8, pp. 26978.
28. W.P. Jones and B.E. Launder: Int. J. Heat Mass Transfer, 1972,
vol. 15, pp. 30114.
29. B.E. Launder and D.B. Spalding: Comput. Math. Appl. Mech.
Eng., 1974, vol. 3, pp. 26989.
30. PHOENICS 2006, D.B. Spalding, CHAM, Bakery House,
London.
31. Esperanza Rodr guez M.: Reporte Interno de Doctorado,
CINVESTAV, Feb. 2006.

METALLURGICAL AND MATERIALS TRANSACTIONS B